<< return to Pixycam.com

Shape Detection

Hey folks!
Just ordered my cam last week and looking forward to having it in my hands soon.

I would just like to have this thing activate a WAV or MP3 trigger when it detects a preset simple shape.

I imagine that the face detection algorithm is very complex, can you also include a very basic shape detector? It seems to me that a simple shape detector could be fairly simple to come up with if complete face detection is on it’s way.

I have seen some of the shapes that other systems use and they store very small black and white pixel block bmp files.

Thanks!!

Hi OED,

Face detection and shape detection can be very different features to implement (or very similar) depending on which algorithm is used. I think as of right now we are going to focus on getting face detection to work well before considering adding on to it.

We have not had enough requests for this to be at the top of our to-do list. I would imagine something like this gets implemented eventually, but as of right now we don’t have plans to.

Scott

Can the face detection detect animal faces?

Could you point me into the right direction so I could start learning about the face detection algorithms that you are using? What would be the easiest way for me to start understanding and possibly toying with this stuff myself? I have installed LPCXpresso on Debian so far… and waiting for my Pixy. :slight_smile:

Thanks!

Shape detection would be extremely useful to many people I think. If shape detection was implemented I could have a completely autonomous system for the Airborne Delivery Challenge! I am using Pixy to recognise Outback Joe and drop the packages. At the moment I need a second camera facing downwards so I can see the “rescue” symbols in the second mission. Unfortunately the competition is in a month so I don’t think shape recognition is going to be available before then… :frowning:

EDIT:
It also appears that a few other people would like this feature too. (http://www.cmucam.org/boards/9/topics/3054)

I believe shape detection could be more useful especially if you think of the combinations of colors and shapes that you can use as triggers.
I could be totally wrong but I think that face detection is an application more suited for a desktop user wanting to video chat not a sensor.

Hi all,

Thanks for the input! We’ll keep it in mind for the next round of features.

OED: As for detecting animal faces, I wouldn’t count on it. Detecting human faces is hard enough, and detecting animals (especially without specifying which kind) would be even harder. I’m sure there are research papers out there on the subject.

As for face detection algorithms, the two I’m most familiar with that you can look up are the Viola-Jones face detector and neural network face detection. However, I’m not actually implementing this feature for Pixy so I don’t know what algorithm is being used. I’ll see if I can get more info.

Scott

Great, thank you very much Scott.

Starting a new thread for Edge Detection - It looks like this would be much easier to do and there is already some code available for the “Video Experimenter shield” but will need to be modified or rewritten.

I bought the Pixy partially on the announcement of facial detection being implemented,so I would hope that stays as a priority.

I bought the Pixy partially on the announcement of facial detection being implemented,so I would hope that stays as a priority. As far as it being a desktop only feature, I disagree. I am build a pet-like robot and I’d like it to recognize a human that is right in front of it. I would be curious if you could switch easily between color and facial recognition. The idea is that when the robot detects a face, it asks for a ball. When one is produced it will chase and kick it.

I personally would like to implement facial recognition on a quadcopter/uav in a swarm set up to quickly identify someone in a crowd